2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
5 #include <linux/config.h>
6 #include <linux/time.h>
8 #include <linux/reiserfs_fs.h>
9 #include <linux/reiserfs_acl.h>
10 #include <linux/reiserfs_xattr.h>
11 #include <linux/smp_lock.h>
12 #include <linux/pagemap.h>
13 #include <linux/highmem.h>
14 #include <asm/uaccess.h>
15 #include <asm/unaligned.h>
16 #include <linux/buffer_head.h>
17 #include <linux/mpage.h>
18 #include <linux/writeback.h>
19 #include <linux/quotaops.h>
20 #include <linux/vserver/xid.h>
22 extern int reiserfs_default_io_size; /* default io size devuned in super.c */
24 /* args for the create parameter of reiserfs_get_block */
25 #define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */
26 #define GET_BLOCK_CREATE 1 /* add anything you need to find block */
27 #define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */
28 #define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */
29 #define GET_BLOCK_NO_ISEM 8 /* i_sem is not held, don't preallocate */
30 #define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */
32 static int reiserfs_get_block (struct inode * inode, sector_t block,
33 struct buffer_head * bh_result, int create);
34 static int reiserfs_commit_write(struct file *f, struct page *page,
35 unsigned from, unsigned to);
37 void reiserfs_delete_inode (struct inode * inode)
39 int jbegin_count = JOURNAL_PER_BALANCE_CNT * 2;
40 struct reiserfs_transaction_handle th ;
42 reiserfs_write_lock(inode->i_sb);
44 DQUOT_FREE_INODE(inode);
45 /* The = 0 happens when we abort creating a new inode for some reason like lack of space.. */
46 if (!(inode->i_state & I_NEW) && INODE_PKEY(inode)->k_objectid != 0) { /* also handles bad_inode case */
49 reiserfs_delete_xattrs (inode);
51 journal_begin(&th, inode->i_sb, jbegin_count) ;
52 reiserfs_update_inode_transaction(inode) ;
54 reiserfs_delete_object (&th, inode);
56 journal_end(&th, inode->i_sb, jbegin_count) ;
60 /* all items of file are deleted, so we can remove "save" link */
61 remove_save_link (inode, 0/* not truncate */);
63 /* no object items are in the tree */
66 clear_inode (inode); /* note this must go after the journal_end to prevent deadlock */
68 reiserfs_write_unlock(inode->i_sb);
71 static void _make_cpu_key (struct cpu_key * key, int version, __u32 dirid, __u32 objectid,
72 loff_t offset, int type, int length )
74 key->version = version;
76 key->on_disk_key.k_dir_id = dirid;
77 key->on_disk_key.k_objectid = objectid;
78 set_cpu_key_k_offset (key, offset);
79 set_cpu_key_k_type (key, type);
80 key->key_length = length;
84 /* take base of inode_key (it comes from inode always) (dirid, objectid) and version from an inode, set
85 offset and type of key */
86 void make_cpu_key (struct cpu_key * key, struct inode * inode, loff_t offset,
87 int type, int length )
89 _make_cpu_key (key, get_inode_item_key_version (inode), le32_to_cpu (INODE_PKEY (inode)->k_dir_id),
90 le32_to_cpu (INODE_PKEY (inode)->k_objectid),
91 offset, type, length);
96 // when key is 0, do not set version and short key
98 inline void make_le_item_head (struct item_head * ih, const struct cpu_key * key,
100 loff_t offset, int type, int length,
101 int entry_count/*or ih_free_space*/)
104 ih->ih_key.k_dir_id = cpu_to_le32 (key->on_disk_key.k_dir_id);
105 ih->ih_key.k_objectid = cpu_to_le32 (key->on_disk_key.k_objectid);
107 put_ih_version( ih, version );
108 set_le_ih_k_offset (ih, offset);
109 set_le_ih_k_type (ih, type);
110 put_ih_item_len( ih, length );
111 /* set_ih_free_space (ih, 0);*/
112 // for directory items it is entry count, for directs and stat
113 // datas - 0xffff, for indirects - 0
114 put_ih_entry_count( ih, entry_count );
118 // FIXME: we might cache recently accessed indirect item
120 // Ugh. Not too eager for that....
121 // I cut the code until such time as I see a convincing argument (benchmark).
122 // I don't want a bloated inode struct..., and I don't like code complexity....
124 /* cutting the code is fine, since it really isn't in use yet and is easy
125 ** to add back in. But, Vladimir has a really good idea here. Think
126 ** about what happens for reading a file. For each page,
127 ** The VFS layer calls reiserfs_readpage, who searches the tree to find
128 ** an indirect item. This indirect item has X number of pointers, where
129 ** X is a big number if we've done the block allocation right. But,
130 ** we only use one or two of these pointers during each call to readpage,
131 ** needlessly researching again later on.
133 ** The size of the cache could be dynamic based on the size of the file.
135 ** I'd also like to see us cache the location the stat data item, since
136 ** we are needlessly researching for that frequently.
141 /* If this page has a file tail in it, and
142 ** it was read in by get_block_create_0, the page data is valid,
143 ** but tail is still sitting in a direct item, and we can't write to
144 ** it. So, look through this page, and check all the mapped buffers
145 ** to make sure they have valid block numbers. Any that don't need
146 ** to be unmapped, so that block_prepare_write will correctly call
147 ** reiserfs_get_block to convert the tail into an unformatted node
149 static inline void fix_tail_page_for_writing(struct page *page) {
150 struct buffer_head *head, *next, *bh ;
152 if (page && page_has_buffers(page)) {
153 head = page_buffers(page) ;
156 next = bh->b_this_page ;
157 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
158 reiserfs_unmap_buffer(bh) ;
161 } while (bh != head) ;
165 /* reiserfs_get_block does not need to allocate a block only if it has been
166 done already or non-hole position has been found in the indirect item */
167 static inline int allocation_needed (int retval, b_blocknr_t allocated,
168 struct item_head * ih,
169 __u32 * item, int pos_in_item)
173 if (retval == POSITION_FOUND && is_indirect_le_ih (ih) &&
174 get_block_num(item, pos_in_item))
179 static inline int indirect_item_found (int retval, struct item_head * ih)
181 return (retval == POSITION_FOUND) && is_indirect_le_ih (ih);
185 static inline void set_block_dev_mapped (struct buffer_head * bh,
186 b_blocknr_t block, struct inode * inode)
188 map_bh(bh, inode->i_sb, block);
193 // files which were created in the earlier version can not be longer,
196 static int file_capable (struct inode * inode, long block)
198 if (get_inode_item_key_version (inode) != KEY_FORMAT_3_5 || // it is new file.
199 block < (1 << (31 - inode->i_sb->s_blocksize_bits))) // old file, but 'block' is inside of 2gb
205 /*static*/ void restart_transaction(struct reiserfs_transaction_handle *th,
206 struct inode *inode, struct path *path) {
207 struct super_block *s = th->t_super ;
208 int len = th->t_blocks_allocated ;
210 /* we cannot restart while nested */
211 if (th->t_refcount > 1) {
215 reiserfs_update_sd(th, inode) ;
216 journal_end(th, s, len) ;
217 journal_begin(th, s, JOURNAL_PER_BALANCE_CNT * 6) ;
218 reiserfs_update_inode_transaction(inode) ;
221 // it is called by get_block when create == 0. Returns block number
222 // for 'block'-th logical block of file. When it hits direct item it
223 // returns 0 (being called from bmap) or read direct item into piece
224 // of page (bh_result)
226 // Please improve the english/clarity in the comment above, as it is
227 // hard to understand.
229 static int _get_block_create_0 (struct inode * inode, long block,
230 struct buffer_head * bh_result,
233 INITIALIZE_PATH (path);
235 struct buffer_head * bh;
236 struct item_head * ih, tmp_ih;
243 unsigned long offset ;
245 // prepare the key to look for the 'block'-th block of file
246 make_cpu_key (&key, inode,
247 (loff_t)block * inode->i_sb->s_blocksize + 1, TYPE_ANY, 3);
250 if (search_for_position_by_key (inode->i_sb, &key, &path) != POSITION_FOUND) {
253 kunmap(bh_result->b_page) ;
254 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
255 // That there is some MMAPED data associated with it that is yet to be written to disk.
256 if ((args & GET_BLOCK_NO_HOLE) && !PageUptodate(bh_result->b_page) ) {
263 bh = get_last_bh (&path);
265 if (is_indirect_le_ih (ih)) {
266 __u32 * ind_item = (__u32 *)B_I_PITEM (bh, ih);
268 /* FIXME: here we could cache indirect item or part of it in
269 the inode to avoid search_by_key in case of subsequent
271 blocknr = get_block_num(ind_item, path.pos_in_item) ;
274 map_bh(bh_result, inode->i_sb, blocknr);
275 if (path.pos_in_item == ((ih_item_len(ih) / UNFM_P_SIZE) - 1)) {
276 set_buffer_boundary(bh_result);
279 // We do not return -ENOENT if there is a hole but page is uptodate, because it means
280 // That there is some MMAPED data associated with it that is yet to be written to disk.
281 if ((args & GET_BLOCK_NO_HOLE) && !PageUptodate(bh_result->b_page) ) {
287 kunmap(bh_result->b_page) ;
291 // requested data are in direct item(s)
292 if (!(args & GET_BLOCK_READ_DIRECT)) {
293 // we are called by bmap. FIXME: we can not map block of file
294 // when it is stored in direct item(s)
297 kunmap(bh_result->b_page) ;
301 /* if we've got a direct item, and the buffer or page was uptodate,
302 ** we don't want to pull data off disk again. skip to the
303 ** end, where we map the buffer and return
305 if (buffer_uptodate(bh_result)) {
309 ** grab_tail_page can trigger calls to reiserfs_get_block on up to date
310 ** pages without any buffers. If the page is up to date, we don't want
311 ** read old data off disk. Set the up to date bit on the buffer instead
312 ** and jump to the end
314 if (!bh_result->b_page || PageUptodate(bh_result->b_page)) {
315 set_buffer_uptodate(bh_result);
319 // read file tail into part of page
320 offset = (cpu_key_k_offset(&key) - 1) & (PAGE_CACHE_SIZE - 1) ;
321 fs_gen = get_generation(inode->i_sb) ;
322 copy_item_head (&tmp_ih, ih);
324 /* we only want to kmap if we are reading the tail into the page.
325 ** this is not the common case, so we don't kmap until we are
326 ** sure we need to. But, this means the item might move if
330 p = (char *)kmap(bh_result->b_page) ;
331 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
336 memset (p, 0, inode->i_sb->s_blocksize);
338 if (!is_direct_le_ih (ih)) {
341 /* make sure we don't read more bytes than actually exist in
342 ** the file. This can happen in odd cases where i_size isn't
343 ** correct, and when direct item padding results in a few
344 ** extra bytes at the end of the direct item
346 if ((le_ih_k_offset(ih) + path.pos_in_item) > inode->i_size)
348 if ((le_ih_k_offset(ih) - 1 + ih_item_len(ih)) > inode->i_size) {
349 chars = inode->i_size - (le_ih_k_offset(ih) - 1) - path.pos_in_item;
352 chars = ih_item_len(ih) - path.pos_in_item;
354 memcpy (p, B_I_PITEM (bh, ih) + path.pos_in_item, chars);
361 if (PATH_LAST_POSITION (&path) != (B_NR_ITEMS (bh) - 1))
362 // we done, if read direct item is not the last item of
363 // node FIXME: we could try to check right delimiting key
364 // to see whether direct item continues in the right
365 // neighbor or rely on i_size
368 // update key to look for the next piece
369 set_cpu_key_k_offset (&key, cpu_key_k_offset (&key) + chars);
370 if (search_for_position_by_key (inode->i_sb, &key, &path) != POSITION_FOUND)
371 // we read something from tail, even if now we got IO_ERROR
373 bh = get_last_bh (&path);
377 flush_dcache_page(bh_result->b_page) ;
378 kunmap(bh_result->b_page) ;
382 /* this buffer has valid data, but isn't valid for io. mapping it to
383 * block #0 tells the rest of reiserfs it just has a tail in it
385 map_bh(bh_result, inode->i_sb, 0);
386 set_buffer_uptodate (bh_result);
391 // this is called to create file map. So, _get_block_create_0 will not
393 int reiserfs_bmap (struct inode * inode, sector_t block,
394 struct buffer_head * bh_result, int create)
396 if (!file_capable (inode, block))
399 reiserfs_write_lock(inode->i_sb);
400 /* do not read the direct item */
401 _get_block_create_0 (inode, block, bh_result, 0) ;
402 reiserfs_write_unlock(inode->i_sb);
406 /* special version of get_block that is only used by grab_tail_page right
407 ** now. It is sent to block_prepare_write, and when you try to get a
408 ** block past the end of the file (or a block from a hole) it returns
409 ** -ENOENT instead of a valid buffer. block_prepare_write expects to
410 ** be able to do i/o on the buffers returned, unless an error value
413 ** So, this allows block_prepare_write to be used for reading a single block
414 ** in a page. Where it does not produce a valid page for holes, or past the
415 ** end of the file. This turns out to be exactly what we need for reading
416 ** tails for conversion.
418 ** The point of the wrapper is forcing a certain value for create, even
419 ** though the VFS layer is calling this function with create==1. If you
420 ** don't want to send create == GET_BLOCK_NO_HOLE to reiserfs_get_block,
421 ** don't use this function.
423 static int reiserfs_get_block_create_0 (struct inode * inode, sector_t block,
424 struct buffer_head * bh_result, int create) {
425 return reiserfs_get_block(inode, block, bh_result, GET_BLOCK_NO_HOLE) ;
428 /* This is special helper for reiserfs_get_block in case we are executing
429 direct_IO request. */
430 static int reiserfs_get_blocks_direct_io(struct inode *inode,
432 unsigned long max_blocks,
433 struct buffer_head *bh_result,
438 bh_result->b_page = NULL;
440 /* We set the b_size before reiserfs_get_block call since it is
441 referenced in convert_tail_for_hole() that may be called from
442 reiserfs_get_block() */
443 bh_result->b_size = (1 << inode->i_blkbits);
445 ret = reiserfs_get_block(inode, iblock, bh_result,
446 create | GET_BLOCK_NO_DANGLE) ;
448 /* don't allow direct io onto tail pages */
449 if (ret == 0 && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
450 /* make sure future calls to the direct io funcs for this offset
451 ** in the file fail by unmapping the buffer
453 clear_buffer_mapped(bh_result);
456 /* Possible unpacked tail. Flush the data before pages have
458 if (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) {
460 reiserfs_commit_for_inode(inode);
461 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
469 ** helper function for when reiserfs_get_block is called for a hole
470 ** but the file tail is still in a direct item
471 ** bh_result is the buffer head for the hole
472 ** tail_offset is the offset of the start of the tail in the file
474 ** This calls prepare_write, which will start a new transaction
475 ** you should not be in a transaction, or have any paths held when you
478 static int convert_tail_for_hole(struct inode *inode,
479 struct buffer_head *bh_result,
480 loff_t tail_offset) {
481 unsigned long index ;
482 unsigned long tail_end ;
483 unsigned long tail_start ;
484 struct page * tail_page ;
485 struct page * hole_page = bh_result->b_page ;
488 if ((tail_offset & (bh_result->b_size - 1)) != 1)
491 /* always try to read until the end of the block */
492 tail_start = tail_offset & (PAGE_CACHE_SIZE - 1) ;
493 tail_end = (tail_start | (bh_result->b_size - 1)) + 1 ;
495 index = tail_offset >> PAGE_CACHE_SHIFT ;
496 /* hole_page can be zero in case of direct_io, we are sure
497 that we cannot get here if we write with O_DIRECT into
499 if (!hole_page || index != hole_page->index) {
500 tail_page = grab_cache_page(inode->i_mapping, index) ;
506 tail_page = hole_page ;
509 /* we don't have to make sure the conversion did not happen while
510 ** we were locking the page because anyone that could convert
511 ** must first take i_sem.
513 ** We must fix the tail page for writing because it might have buffers
514 ** that are mapped, but have a block number of 0. This indicates tail
515 ** data that has been read directly into the page, and block_prepare_write
516 ** won't trigger a get_block in this case.
518 fix_tail_page_for_writing(tail_page) ;
519 retval = reiserfs_prepare_write(NULL, tail_page, tail_start, tail_end);
523 /* tail conversion might change the data in the page */
524 flush_dcache_page(tail_page) ;
526 retval = reiserfs_commit_write(NULL, tail_page, tail_start, tail_end) ;
529 if (tail_page != hole_page) {
530 unlock_page(tail_page) ;
531 page_cache_release(tail_page) ;
537 static inline int _allocate_block(struct reiserfs_transaction_handle *th,
540 b_blocknr_t *allocated_block_nr,
544 #ifdef REISERFS_PREALLOCATE
545 if (!(flags & GET_BLOCK_NO_ISEM)) {
546 return reiserfs_new_unf_blocknrs2(th, inode, allocated_block_nr, path, block);
549 return reiserfs_new_unf_blocknrs (th, inode, allocated_block_nr, path, block);
552 int reiserfs_get_block (struct inode * inode, sector_t block,
553 struct buffer_head * bh_result, int create)
556 b_blocknr_t allocated_block_nr = 0;// b_blocknr_t is (unsigned) 32 bit int
557 INITIALIZE_PATH(path);
560 struct buffer_head * bh, * unbh = NULL;
561 struct item_head * ih, tmp_ih;
565 struct reiserfs_transaction_handle *th = NULL;
566 /* space reserved in transaction batch:
567 . 3 balancings in direct->indirect conversion
568 . 1 block involved into reiserfs_update_sd()
569 XXX in practically impossible worst case direct2indirect()
570 can incur (much) more that 3 balancings. */
571 int jbegin_count = JOURNAL_PER_BALANCE_CNT * 3 + 1;
574 loff_t new_offset = (((loff_t)block) << inode->i_sb->s_blocksize_bits) + 1 ;
577 reiserfs_write_lock(inode->i_sb);
578 version = get_inode_item_key_version (inode);
581 reiserfs_write_unlock(inode->i_sb);
585 if (!file_capable (inode, block)) {
586 reiserfs_write_unlock(inode->i_sb);
590 /* if !create, we aren't changing the FS, so we don't need to
591 ** log anything, so we don't need to start a transaction
593 if (!(create & GET_BLOCK_CREATE)) {
595 /* find number of block-th logical block of the file */
596 ret = _get_block_create_0 (inode, block, bh_result,
597 create | GET_BLOCK_READ_DIRECT) ;
598 reiserfs_write_unlock(inode->i_sb);
602 * if we're already in a transaction, make sure to close
603 * any new transactions we start in this func
605 if ((create & GET_BLOCK_NO_DANGLE) ||
606 reiserfs_transaction_running(inode->i_sb))
609 /* If file is of such a size, that it might have a tail and tails are enabled
610 ** we should mark it as possibly needing tail packing on close
612 if ( (have_large_tails (inode->i_sb) && inode->i_size < i_block_size (inode)*4) ||
613 (have_small_tails (inode->i_sb) && inode->i_size < i_block_size(inode)) )
614 REISERFS_I(inode)->i_flags |= i_pack_on_close_mask ;
616 /* set the key of the first byte in the 'block'-th block of file */
617 make_cpu_key (&key, inode, new_offset,
618 TYPE_ANY, 3/*key length*/);
619 if ((new_offset + inode->i_sb->s_blocksize - 1) > inode->i_size) {
621 th = reiserfs_persistent_transaction(inode->i_sb, jbegin_count);
626 reiserfs_update_inode_transaction(inode) ;
630 retval = search_for_position_by_key (inode->i_sb, &key, &path);
631 if (retval == IO_ERROR) {
636 bh = get_last_bh (&path);
638 item = get_item (&path);
639 pos_in_item = path.pos_in_item;
641 fs_gen = get_generation (inode->i_sb);
642 copy_item_head (&tmp_ih, ih);
644 if (allocation_needed (retval, allocated_block_nr, ih, item, pos_in_item)) {
645 /* we have to allocate block for the unformatted node */
651 repeat = _allocate_block(th, block, inode, &allocated_block_nr, &path, create);
653 if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
654 /* restart the transaction to give the journal a chance to free
655 ** some blocks. releases the path, so we have to go back to
656 ** research if we succeed on the second try
658 SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;
659 restart_transaction(th, inode, &path) ;
660 repeat = _allocate_block(th, block, inode, &allocated_block_nr, NULL, create);
662 if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
665 if (repeat == QUOTA_EXCEEDED)
672 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
677 if (indirect_item_found (retval, ih)) {
678 b_blocknr_t unfm_ptr;
679 /* 'block'-th block is in the file already (there is
680 corresponding cell in some indirect item). But it may be
681 zero unformatted node pointer (hole) */
682 unfm_ptr = get_block_num (item, pos_in_item);
684 /* use allocated block to plug the hole */
685 reiserfs_prepare_for_journal(inode->i_sb, bh, 1) ;
686 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
687 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
690 set_buffer_new(bh_result);
691 if (buffer_dirty(bh_result) && reiserfs_data_ordered(inode->i_sb))
692 reiserfs_add_ordered_list(inode, bh_result);
693 put_block_num(item, pos_in_item, allocated_block_nr) ;
694 unfm_ptr = allocated_block_nr;
695 journal_mark_dirty (th, inode->i_sb, bh);
696 reiserfs_update_sd(th, inode) ;
698 set_block_dev_mapped(bh_result, unfm_ptr, inode);
701 reiserfs_end_persistent_transaction(th);
703 reiserfs_write_unlock(inode->i_sb);
705 /* the item was found, so new blocks were not added to the file
706 ** there is no need to make sure the inode is updated with this
717 /* desired position is not found or is in the direct item. We have
718 to append file with holes up to 'block'-th block converting
719 direct items to indirect one if necessary */
722 if (is_statdata_le_ih (ih)) {
724 struct cpu_key tmp_key;
726 /* indirect item has to be inserted */
727 make_le_item_head (&tmp_ih, &key, version, 1, TYPE_INDIRECT,
728 UNFM_P_SIZE, 0/* free_space */);
730 if (cpu_key_k_offset (&key) == 1) {
731 /* we are going to add 'block'-th block to the file. Use
732 allocated block for that */
733 unp = cpu_to_le32 (allocated_block_nr);
734 set_block_dev_mapped (bh_result, allocated_block_nr, inode);
735 set_buffer_new(bh_result);
739 set_cpu_key_k_offset (&tmp_key, 1);
740 PATH_LAST_POSITION(&path) ++;
742 retval = reiserfs_insert_item (th, &path, &tmp_key, &tmp_ih, inode, (char *)&unp);
744 reiserfs_free_block (th, inode, allocated_block_nr, 1);
745 goto failure; // retval == -ENOSPC, -EDQUOT or -EIO or -EEXIST
747 //mark_tail_converted (inode);
748 } else if (is_direct_le_ih (ih)) {
749 /* direct item has to be converted */
752 tail_offset = ((le_ih_k_offset (ih) - 1) & ~(inode->i_sb->s_blocksize - 1)) + 1;
753 if (tail_offset == cpu_key_k_offset (&key)) {
754 /* direct item we just found fits into block we have
755 to map. Convert it into unformatted node: use
756 bh_result for the conversion */
757 set_block_dev_mapped (bh_result, allocated_block_nr, inode);
761 /* we have to padd file tail stored in direct item(s)
762 up to block size and convert it to unformatted
763 node. FIXME: this should also get into page cache */
767 * ugly, but we can only end the transaction if
770 if (th->t_refcount == 1) {
771 reiserfs_end_persistent_transaction(th);
775 retval = convert_tail_for_hole(inode, bh_result, tail_offset) ;
777 if ( retval != -ENOSPC )
778 reiserfs_warning (inode->i_sb, "clm-6004: convert tail failed inode %lu, error %d", inode->i_ino, retval) ;
779 if (allocated_block_nr) {
780 /* the bitmap, the super, and the stat data == 3 */
782 th = reiserfs_persistent_transaction(inode->i_sb,3);
784 reiserfs_free_block (th,inode,allocated_block_nr,1);
790 retval = direct2indirect (th, inode, &path, unbh, tail_offset);
792 reiserfs_unmap_buffer(unbh);
793 reiserfs_free_block (th, inode, allocated_block_nr, 1);
796 /* it is important the set_buffer_uptodate is done after
797 ** the direct2indirect. The buffer might contain valid
798 ** data newer than the data on disk (read by readpage, changed,
799 ** and then sent here by writepage). direct2indirect needs
800 ** to know if unbh was already up to date, so it can decide
801 ** if the data in unbh needs to be replaced with data from
804 set_buffer_uptodate (unbh);
806 /* unbh->b_page == NULL in case of DIRECT_IO request, this means
807 buffer will disappear shortly, so it should not be added to
809 if ( unbh->b_page ) {
810 /* we've converted the tail, so we must
811 ** flush unbh before the transaction commits
813 reiserfs_add_tail_list(inode, unbh) ;
815 /* mark it dirty now to prevent commit_write from adding
816 ** this buffer to the inode's dirty buffer list
819 * AKPM: changed __mark_buffer_dirty to mark_buffer_dirty().
820 * It's still atomic, but it sets the page dirty too,
821 * which makes it eligible for writeback at any time by the
822 * VM (which was also the case with __mark_buffer_dirty())
824 mark_buffer_dirty(unbh) ;
827 /* append indirect item with holes if needed, when appending
828 pointer to 'block'-th block use block, which is already
830 struct cpu_key tmp_key;
831 unp_t unf_single=0; // We use this in case we need to allocate only
832 // one block which is a fastpath
834 __u64 max_to_insert=MAX_ITEM_LEN(inode->i_sb->s_blocksize)/UNFM_P_SIZE;
837 RFALSE( pos_in_item != ih_item_len(ih) / UNFM_P_SIZE,
838 "vs-804: invalid position for append");
839 /* indirect item has to be appended, set up key of that position */
840 make_cpu_key (&tmp_key, inode,
841 le_key_k_offset (version, &(ih->ih_key)) + op_bytes_number (ih, inode->i_sb->s_blocksize),
842 //pos_in_item * inode->i_sb->s_blocksize,
843 TYPE_INDIRECT, 3);// key type is unimportant
845 blocks_needed = 1 + ((cpu_key_k_offset (&key) - cpu_key_k_offset (&tmp_key)) >> inode->i_sb->s_blocksize_bits);
846 RFALSE( blocks_needed < 0, "green-805: invalid offset");
848 if ( blocks_needed == 1 ) {
851 un=kmalloc( min(blocks_needed,max_to_insert)*UNFM_P_SIZE,
852 GFP_ATOMIC); // We need to avoid scheduling.
858 memset(un, 0, UNFM_P_SIZE * min(blocks_needed,max_to_insert));
860 if ( blocks_needed <= max_to_insert) {
861 /* we are going to add target block to the file. Use allocated
863 un[blocks_needed-1] = cpu_to_le32 (allocated_block_nr);
864 set_block_dev_mapped (bh_result, allocated_block_nr, inode);
865 set_buffer_new(bh_result);
868 /* paste hole to the indirect item */
869 /* If kmalloc failed, max_to_insert becomes zero and it means we
870 only have space for one block */
871 blocks_needed=max_to_insert?max_to_insert:1;
873 retval = reiserfs_paste_into_item (th, &path, &tmp_key, inode, (char *)un, UNFM_P_SIZE * blocks_needed);
875 if (blocks_needed != 1)
879 reiserfs_free_block (th, inode, allocated_block_nr, 1);
883 /* We need to mark new file size in case this function will be
884 interrupted/aborted later on. And we may do this only for
886 inode->i_size += inode->i_sb->s_blocksize * blocks_needed;
893 /* this loop could log more blocks than we had originally asked
894 ** for. So, we have to allow the transaction to end if it is
895 ** too big or too full. Update the inode so things are
896 ** consistent if we crash before the function returns
898 ** release the path so that anybody waiting on the path before
899 ** ending their transaction will be able to continue.
901 if (journal_transaction_should_end(th, th->t_blocks_allocated)) {
902 restart_transaction(th, inode, &path) ;
904 /* inserting indirect pointers for a hole can take a
905 ** long time. reschedule if needed
909 retval = search_for_position_by_key (inode->i_sb, &key, &path);
910 if (retval == IO_ERROR) {
914 if (retval == POSITION_FOUND) {
915 reiserfs_warning (inode->i_sb, "vs-825: reiserfs_get_block: "
916 "%K should not be found", &key);
918 if (allocated_block_nr)
919 reiserfs_free_block (th, inode, allocated_block_nr, 1);
923 bh = get_last_bh (&path);
925 item = get_item (&path);
926 pos_in_item = path.pos_in_item;
934 reiserfs_update_sd(th, inode) ;
935 reiserfs_end_persistent_transaction(th);
937 reiserfs_write_unlock(inode->i_sb);
938 reiserfs_check_path(&path) ;
943 reiserfs_readpages(struct file *file, struct address_space *mapping,
944 struct list_head *pages, unsigned nr_pages)
946 return mpage_readpages(mapping, pages, nr_pages, reiserfs_get_block);
949 /* Compute real number of used bytes by file
950 * Following three functions can go away when we'll have enough space in stat item
952 static int real_space_diff(struct inode *inode, int sd_size)
955 loff_t blocksize = inode->i_sb->s_blocksize ;
957 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode))
960 /* End of file is also in full block with indirect reference, so round
961 ** up to the next block.
963 ** there is just no way to know if the tail is actually packed
964 ** on the file, so we have to assume it isn't. When we pack the
965 ** tail, we add 4 bytes to pretend there really is an unformatted
968 bytes = ((inode->i_size + (blocksize-1)) >> inode->i_sb->s_blocksize_bits) * UNFM_P_SIZE + sd_size;
972 static inline loff_t to_real_used_space(struct inode *inode, ulong blocks,
975 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
976 return inode->i_size + (loff_t)(real_space_diff(inode, sd_size)) ;
978 return ((loff_t)real_space_diff(inode, sd_size)) + (((loff_t)blocks) << 9);
981 /* Compute number of blocks used by file in ReiserFS counting */
982 static inline ulong to_fake_used_blocks(struct inode *inode, int sd_size)
984 loff_t bytes = inode_get_bytes(inode) ;
985 loff_t real_space = real_space_diff(inode, sd_size) ;
987 /* keeps fsck and non-quota versions of reiserfs happy */
988 if (S_ISLNK(inode->i_mode) || S_ISDIR(inode->i_mode)) {
989 bytes += (loff_t)511 ;
992 /* files from before the quota patch might i_blocks such that
993 ** bytes < real_space. Deal with that here to prevent it from
996 if (bytes < real_space)
998 return (bytes - real_space) >> 9;
1002 // BAD: new directories have stat data of new type and all other items
1003 // of old type. Version stored in the inode says about body items, so
1004 // in update_stat_data we can not rely on inode, but have to check
1005 // item version directly
1008 // called by read_locked_inode
1009 static void init_inode (struct inode * inode, struct path * path)
1011 struct buffer_head * bh;
1012 struct item_head * ih;
1016 //int version = ITEM_VERSION_1;
1018 bh = PATH_PLAST_BUFFER (path);
1019 ih = PATH_PITEM_HEAD (path);
1022 copy_key (INODE_PKEY (inode), &(ih->ih_key));
1023 inode->i_blksize = reiserfs_default_io_size;
1025 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list ));
1026 REISERFS_I(inode)->i_flags = 0;
1027 REISERFS_I(inode)->i_prealloc_block = 0;
1028 REISERFS_I(inode)->i_prealloc_count = 0;
1029 REISERFS_I(inode)->i_trans_id = 0;
1030 REISERFS_I(inode)->i_jl = NULL;
1031 REISERFS_I(inode)->i_acl_access = NULL;
1032 REISERFS_I(inode)->i_acl_default = NULL;
1033 init_rwsem (&REISERFS_I(inode)->xattr_sem);
1035 if (stat_data_v1 (ih)) {
1036 struct stat_data_v1 * sd = (struct stat_data_v1 *)B_I_PITEM (bh, ih);
1037 unsigned long blocks;
1039 uid = sd_v1_uid(sd);
1040 gid = sd_v1_gid(sd);
1042 set_inode_item_key_version (inode, KEY_FORMAT_3_5);
1043 set_inode_sd_version (inode, STAT_DATA_V1);
1044 inode->i_mode = sd_v1_mode(sd);
1045 inode->i_nlink = sd_v1_nlink(sd);
1046 inode->i_size = sd_v1_size(sd);
1047 inode->i_atime.tv_sec = sd_v1_atime(sd);
1048 inode->i_mtime.tv_sec = sd_v1_mtime(sd);
1049 inode->i_ctime.tv_sec = sd_v1_ctime(sd);
1050 inode->i_atime.tv_nsec = 0;
1051 inode->i_ctime.tv_nsec = 0;
1052 inode->i_mtime.tv_nsec = 0;
1054 inode->i_blocks = sd_v1_blocks(sd);
1055 inode->i_generation = le32_to_cpu (INODE_PKEY (inode)->k_dir_id);
1056 blocks = (inode->i_size + 511) >> 9;
1057 blocks = _ROUND_UP (blocks, inode->i_sb->s_blocksize >> 9);
1058 if (inode->i_blocks > blocks) {
1059 // there was a bug in <=3.5.23 when i_blocks could take negative
1060 // values. Starting from 3.5.17 this value could even be stored in
1061 // stat data. For such files we set i_blocks based on file
1062 // size. Just 2 notes: this can be wrong for sparce files. On-disk value will be
1063 // only updated if file's inode will ever change
1064 inode->i_blocks = blocks;
1067 rdev = sd_v1_rdev(sd);
1068 REISERFS_I(inode)->i_first_direct_byte = sd_v1_first_direct_byte(sd);
1069 /* an early bug in the quota code can give us an odd number for the
1070 ** block count. This is incorrect, fix it here.
1072 if (inode->i_blocks & 1) {
1075 inode_set_bytes(inode, to_real_used_space(inode, inode->i_blocks,
1077 /* nopack is initially zero for v1 objects. For v2 objects,
1078 nopack is initialised from sd_attrs */
1079 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
1081 // new stat data found, but object may have old items
1082 // (directories and symlinks)
1083 struct stat_data * sd = (struct stat_data *)B_I_PITEM (bh, ih);
1085 uid = sd_v2_uid(sd);
1086 gid = sd_v2_gid(sd);
1088 inode->i_mode = sd_v2_mode(sd);
1089 inode->i_nlink = sd_v2_nlink(sd);
1090 inode->i_size = sd_v2_size(sd);
1091 inode->i_mtime.tv_sec = sd_v2_mtime(sd);
1092 inode->i_atime.tv_sec = sd_v2_atime(sd);
1093 inode->i_ctime.tv_sec = sd_v2_ctime(sd);
1094 inode->i_ctime.tv_nsec = 0;
1095 inode->i_mtime.tv_nsec = 0;
1096 inode->i_atime.tv_nsec = 0;
1097 inode->i_blocks = sd_v2_blocks(sd);
1098 rdev = sd_v2_rdev(sd);
1099 if( S_ISCHR( inode -> i_mode ) || S_ISBLK( inode -> i_mode ) )
1100 inode->i_generation = le32_to_cpu (INODE_PKEY (inode)->k_dir_id);
1102 inode->i_generation = sd_v2_generation(sd);
1104 if (S_ISDIR (inode->i_mode) || S_ISLNK (inode->i_mode))
1105 set_inode_item_key_version (inode, KEY_FORMAT_3_5);
1107 set_inode_item_key_version (inode, KEY_FORMAT_3_6);
1108 REISERFS_I(inode)->i_first_direct_byte = 0;
1109 set_inode_sd_version (inode, STAT_DATA_V2);
1110 inode_set_bytes(inode, to_real_used_space(inode, inode->i_blocks,
1112 /* read persistent inode attributes from sd and initalise
1113 generic inode flags from them */
1114 REISERFS_I(inode)->i_attrs = sd_v2_attrs( sd );
1115 sd_attrs_to_i_attrs( sd_v2_attrs( sd ), inode );
1117 inode->i_uid = INOXID_UID(XID_TAG(inode), uid, gid);
1118 inode->i_gid = INOXID_GID(XID_TAG(inode), uid, gid);
1119 inode->i_xid = INOXID_XID(XID_TAG(inode), uid, gid, 0);
1122 if (S_ISREG (inode->i_mode)) {
1123 inode->i_op = &reiserfs_file_inode_operations;
1124 inode->i_fop = &reiserfs_file_operations;
1125 inode->i_mapping->a_ops = &reiserfs_address_space_operations ;
1126 } else if (S_ISDIR (inode->i_mode)) {
1127 inode->i_op = &reiserfs_dir_inode_operations;
1128 inode->i_fop = &reiserfs_dir_operations;
1129 } else if (S_ISLNK (inode->i_mode)) {
1130 inode->i_op = &reiserfs_symlink_inode_operations;
1131 inode->i_mapping->a_ops = &reiserfs_address_space_operations;
1133 inode->i_blocks = 0;
1134 inode->i_op = &reiserfs_special_inode_operations;
1135 init_special_inode(inode, inode->i_mode, new_decode_dev(rdev));
1140 // update new stat data with inode fields
1141 static void inode2sd (void * sd, struct inode * inode, loff_t size)
1143 struct stat_data * sd_v2 = (struct stat_data *)sd;
1144 uid_t uid = XIDINO_UID(XID_TAG(inode), inode->i_uid, inode->i_xid);
1145 gid_t gid = XIDINO_GID(XID_TAG(inode), inode->i_gid, inode->i_xid);
1148 set_sd_v2_uid(sd_v2, uid );
1149 set_sd_v2_gid(sd_v2, gid );
1150 set_sd_v2_mode(sd_v2, inode->i_mode );
1151 set_sd_v2_nlink(sd_v2, inode->i_nlink );
1152 set_sd_v2_size(sd_v2, size );
1153 set_sd_v2_mtime(sd_v2, inode->i_mtime.tv_sec );
1154 set_sd_v2_atime(sd_v2, inode->i_atime.tv_sec );
1155 set_sd_v2_ctime(sd_v2, inode->i_ctime.tv_sec );
1156 set_sd_v2_blocks(sd_v2, to_fake_used_blocks(inode, SD_V2_SIZE));
1157 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1158 set_sd_v2_rdev(sd_v2, new_encode_dev(inode->i_rdev));
1160 set_sd_v2_generation(sd_v2, inode->i_generation);
1161 flags = REISERFS_I(inode)->i_attrs;
1162 i_attrs_to_sd_attrs( inode, &flags );
1163 set_sd_v2_attrs( sd_v2, flags );
1167 // used to copy inode's fields to old stat data
1168 static void inode2sd_v1 (void * sd, struct inode * inode, loff_t size)
1170 struct stat_data_v1 * sd_v1 = (struct stat_data_v1 *)sd;
1172 set_sd_v1_mode(sd_v1, inode->i_mode );
1173 set_sd_v1_uid(sd_v1, inode->i_uid );
1174 set_sd_v1_gid(sd_v1, inode->i_gid );
1175 set_sd_v1_nlink(sd_v1, inode->i_nlink );
1176 set_sd_v1_size(sd_v1, size );
1177 set_sd_v1_atime(sd_v1, inode->i_atime.tv_sec );
1178 set_sd_v1_ctime(sd_v1, inode->i_ctime.tv_sec );
1179 set_sd_v1_mtime(sd_v1, inode->i_mtime.tv_sec );
1181 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode))
1182 set_sd_v1_rdev(sd_v1, new_encode_dev(inode->i_rdev));
1184 set_sd_v1_blocks(sd_v1, to_fake_used_blocks(inode, SD_V1_SIZE));
1186 // Sigh. i_first_direct_byte is back
1187 set_sd_v1_first_direct_byte(sd_v1, REISERFS_I(inode)->i_first_direct_byte);
1191 /* NOTE, you must prepare the buffer head before sending it here,
1192 ** and then log it after the call
1194 static void update_stat_data (struct path * path, struct inode * inode,
1197 struct buffer_head * bh;
1198 struct item_head * ih;
1200 bh = PATH_PLAST_BUFFER (path);
1201 ih = PATH_PITEM_HEAD (path);
1203 if (!is_statdata_le_ih (ih))
1204 reiserfs_panic (inode->i_sb, "vs-13065: update_stat_data: key %k, found item %h",
1205 INODE_PKEY (inode), ih);
1207 if (stat_data_v1 (ih)) {
1208 // path points to old stat data
1209 inode2sd_v1 (B_I_PITEM (bh, ih), inode, size);
1211 inode2sd (B_I_PITEM (bh, ih), inode, size);
1218 void reiserfs_update_sd_size (struct reiserfs_transaction_handle *th,
1219 struct inode * inode, loff_t size)
1222 INITIALIZE_PATH(path);
1223 struct buffer_head *bh ;
1225 struct item_head *ih, tmp_ih ;
1228 make_cpu_key (&key, inode, SD_OFFSET, TYPE_STAT_DATA, 3);//key type is unimportant
1232 /* look for the object's stat data */
1233 retval = search_item (inode->i_sb, &key, &path);
1234 if (retval == IO_ERROR) {
1235 reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: "
1236 "i/o failure occurred trying to update %K stat data",
1240 if (retval == ITEM_NOT_FOUND) {
1241 pos = PATH_LAST_POSITION (&path);
1243 if (inode->i_nlink == 0) {
1244 /*reiserfs_warning (inode->i_sb, "vs-13050: reiserfs_update_sd: i_nlink == 0, stat data not found");*/
1247 reiserfs_warning (inode->i_sb, "vs-13060: reiserfs_update_sd: "
1248 "stat data of object %k (nlink == %d) not found (pos %d)",
1249 INODE_PKEY (inode), inode->i_nlink, pos);
1250 reiserfs_check_path(&path) ;
1254 /* sigh, prepare_for_journal might schedule. When it schedules the
1255 ** FS might change. We have to detect that, and loop back to the
1256 ** search if the stat data item has moved
1258 bh = get_last_bh(&path) ;
1259 ih = get_ih(&path) ;
1260 copy_item_head (&tmp_ih, ih);
1261 fs_gen = get_generation (inode->i_sb);
1262 reiserfs_prepare_for_journal(inode->i_sb, bh, 1) ;
1263 if (fs_changed (fs_gen, inode->i_sb) && item_moved(&tmp_ih, &path)) {
1264 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
1265 continue ; /* Stat_data item has been moved after scheduling. */
1269 update_stat_data (&path, inode, size);
1270 journal_mark_dirty(th, th->t_super, bh) ;
1275 /* reiserfs_read_locked_inode is called to read the inode off disk, and it
1276 ** does a make_bad_inode when things go wrong. But, we need to make sure
1277 ** and clear the key in the private portion of the inode, otherwise a
1278 ** corresponding iput might try to delete whatever object the inode last
1281 static void reiserfs_make_bad_inode(struct inode *inode) {
1282 memset(INODE_PKEY(inode), 0, KEY_SIZE);
1283 make_bad_inode(inode);
1287 // initially this function was derived from minix or ext2's analog and
1288 // evolved as the prototype did
1291 int reiserfs_init_locked_inode (struct inode * inode, void *p)
1293 struct reiserfs_iget_args *args = (struct reiserfs_iget_args *)p ;
1294 inode->i_ino = args->objectid;
1295 INODE_PKEY(inode)->k_dir_id = cpu_to_le32(args->dirid);
1299 /* looks for stat data in the tree, and fills up the fields of in-core
1300 inode stat data fields */
1301 void reiserfs_read_locked_inode (struct inode * inode, struct reiserfs_iget_args *args)
1303 INITIALIZE_PATH (path_to_sd);
1305 unsigned long dirino;
1308 dirino = args->dirid ;
1310 /* set version 1, version 2 could be used too, because stat data
1311 key is the same in both versions */
1312 key.version = KEY_FORMAT_3_5;
1313 key.on_disk_key.k_dir_id = dirino;
1314 key.on_disk_key.k_objectid = inode->i_ino;
1315 key.on_disk_key.u.k_offset_v1.k_offset = SD_OFFSET;
1316 key.on_disk_key.u.k_offset_v1.k_uniqueness = SD_UNIQUENESS;
1318 /* look for the object's stat data */
1319 retval = search_item (inode->i_sb, &key, &path_to_sd);
1320 if (retval == IO_ERROR) {
1321 reiserfs_warning (inode->i_sb, "vs-13070: reiserfs_read_locked_inode: "
1322 "i/o failure occurred trying to find stat data of %K",
1324 reiserfs_make_bad_inode(inode) ;
1327 if (retval != ITEM_FOUND) {
1328 /* a stale NFS handle can trigger this without it being an error */
1329 pathrelse (&path_to_sd);
1330 reiserfs_make_bad_inode(inode) ;
1335 init_inode (inode, &path_to_sd);
1337 /* It is possible that knfsd is trying to access inode of a file
1338 that is being removed from the disk by some other thread. As we
1339 update sd on unlink all that is required is to check for nlink
1340 here. This bug was first found by Sizif when debugging
1341 SquidNG/Butterfly, forgotten, and found again after Philippe
1342 Gramoulle <philippe.gramoulle@mmania.com> reproduced it.
1344 More logical fix would require changes in fs/inode.c:iput() to
1345 remove inode from hash-table _after_ fs cleaned disk stuff up and
1346 in iget() to return NULL if I_FREEING inode is found in
1348 /* Currently there is one place where it's ok to meet inode with
1349 nlink==0: processing of open-unlinked and half-truncated files
1350 during mount (fs/reiserfs/super.c:finish_unfinished()). */
1351 if( ( inode -> i_nlink == 0 ) &&
1352 ! REISERFS_SB(inode -> i_sb) -> s_is_unlinked_ok ) {
1353 reiserfs_warning (inode->i_sb,
1354 "vs-13075: reiserfs_read_locked_inode: "
1355 "dead inode read from disk %K. "
1356 "This is likely to be race with knfsd. Ignore",
1358 reiserfs_make_bad_inode( inode );
1361 reiserfs_check_path(&path_to_sd) ; /* init inode should be relsing */
1366 * reiserfs_find_actor() - "find actor" reiserfs supplies to iget5_locked().
1368 * @inode: inode from hash table to check
1369 * @opaque: "cookie" passed to iget5_locked(). This is &reiserfs_iget_args.
1371 * This function is called by iget5_locked() to distinguish reiserfs inodes
1372 * having the same inode numbers. Such inodes can only exist due to some
1373 * error condition. One of them should be bad. Inodes with identical
1374 * inode numbers (objectids) are distinguished by parent directory ids.
1377 int reiserfs_find_actor( struct inode *inode, void *opaque )
1379 struct reiserfs_iget_args *args;
1382 /* args is already in CPU order */
1383 return (inode->i_ino == args->objectid) &&
1384 (le32_to_cpu(INODE_PKEY(inode)->k_dir_id) == args->dirid);
1387 struct inode * reiserfs_iget (struct super_block * s, const struct cpu_key * key)
1389 struct inode * inode;
1390 struct reiserfs_iget_args args ;
1392 args.objectid = key->on_disk_key.k_objectid ;
1393 args.dirid = key->on_disk_key.k_dir_id ;
1394 inode = iget5_locked (s, key->on_disk_key.k_objectid,
1395 reiserfs_find_actor, reiserfs_init_locked_inode, (void *)(&args));
1397 return ERR_PTR(-ENOMEM) ;
1399 if (inode->i_state & I_NEW) {
1400 reiserfs_read_locked_inode(inode, &args);
1401 unlock_new_inode(inode);
1404 if (comp_short_keys (INODE_PKEY (inode), key) || is_bad_inode (inode)) {
1405 /* either due to i/o error or a stale NFS handle */
1412 struct dentry *reiserfs_get_dentry(struct super_block *sb, void *vobjp)
1414 __u32 *data = vobjp;
1415 struct cpu_key key ;
1416 struct dentry *result;
1417 struct inode *inode;
1419 key.on_disk_key.k_objectid = data[0] ;
1420 key.on_disk_key.k_dir_id = data[1] ;
1421 inode = reiserfs_iget(sb, &key) ;
1422 if (inode && !IS_ERR(inode) && data[2] != 0 &&
1423 data[2] != inode->i_generation) {
1428 inode = ERR_PTR(-ESTALE);
1430 return ERR_PTR(PTR_ERR(inode));
1431 result = d_alloc_anon(inode);
1434 return ERR_PTR(-ENOMEM);
1439 struct dentry *reiserfs_decode_fh(struct super_block *sb, __u32 *data,
1440 int len, int fhtype,
1441 int (*acceptable)(void *contect, struct dentry *de),
1443 __u32 obj[3], parent[3];
1445 /* fhtype happens to reflect the number of u32s encoded.
1446 * due to a bug in earlier code, fhtype might indicate there
1447 * are more u32s then actually fitted.
1448 * so if fhtype seems to be more than len, reduce fhtype.
1450 * 2 - objectid + dir_id - legacy support
1451 * 3 - objectid + dir_id + generation
1452 * 4 - objectid + dir_id + objectid and dirid of parent - legacy
1453 * 5 - objectid + dir_id + generation + objectid and dirid of parent
1454 * 6 - as above plus generation of directory
1455 * 6 does not fit in NFSv2 handles
1458 if (fhtype != 6 || len != 5)
1459 reiserfs_warning (sb, "nfsd/reiserfs, fhtype=%d, len=%d - odd",
1466 if (fhtype == 3 || fhtype >= 5)
1468 else obj[2] = 0; /* generation number */
1471 parent[0] = data[fhtype>=5?3:2] ;
1472 parent[1] = data[fhtype>=5?4:3] ;
1474 parent[2] = data[5];
1477 return sb->s_export_op->find_exported_dentry(sb, obj, fhtype < 4 ? NULL : parent,
1478 acceptable, context);
1481 int reiserfs_encode_fh(struct dentry *dentry, __u32 *data, int *lenp, int need_parent) {
1482 struct inode *inode = dentry->d_inode ;
1488 data[0] = inode->i_ino ;
1489 data[1] = le32_to_cpu(INODE_PKEY (inode)->k_dir_id) ;
1490 data[2] = inode->i_generation ;
1492 /* no room for directory info? return what we've stored so far */
1493 if (maxlen < 5 || ! need_parent)
1496 spin_lock(&dentry->d_lock);
1497 inode = dentry->d_parent->d_inode ;
1498 data[3] = inode->i_ino ;
1499 data[4] = le32_to_cpu(INODE_PKEY (inode)->k_dir_id) ;
1502 data[5] = inode->i_generation ;
1505 spin_unlock(&dentry->d_lock);
1510 /* looks for stat data, then copies fields to it, marks the buffer
1511 containing stat data as dirty */
1512 /* reiserfs inodes are never really dirty, since the dirty inode call
1513 ** always logs them. This call allows the VFS inode marking routines
1514 ** to properly mark inodes for datasync and such, but only actually
1515 ** does something when called for a synchronous update.
1517 int reiserfs_write_inode (struct inode * inode, int do_sync) {
1518 struct reiserfs_transaction_handle th ;
1519 int jbegin_count = 1 ;
1521 if (inode->i_sb->s_flags & MS_RDONLY) {
1522 reiserfs_warning (inode->i_sb,
1523 "clm-6005: writing inode %lu on readonly FS",
1527 /* memory pressure can sometimes initiate write_inode calls with sync == 1,
1528 ** these cases are just when the system needs ram, not when the
1529 ** inode needs to reach disk for safety, and they can safely be
1530 ** ignored because the altered inode has already been logged.
1532 if (do_sync && !(current->flags & PF_MEMALLOC)) {
1533 reiserfs_write_lock(inode->i_sb);
1534 journal_begin(&th, inode->i_sb, jbegin_count) ;
1535 reiserfs_update_sd (&th, inode);
1536 journal_end_sync(&th, inode->i_sb, jbegin_count) ;
1537 reiserfs_write_unlock(inode->i_sb);
1542 /* FIXME: no need any more. right? */
1543 int reiserfs_sync_inode (struct reiserfs_transaction_handle *th, struct inode * inode)
1547 reiserfs_update_sd (th, inode);
1552 /* stat data of new object is inserted already, this inserts the item
1553 containing "." and ".." entries */
1554 static int reiserfs_new_directory (struct reiserfs_transaction_handle *th,
1555 struct inode *inode,
1556 struct item_head * ih, struct path * path,
1559 struct super_block * sb = th->t_super;
1560 char empty_dir [EMPTY_DIR_SIZE];
1561 char * body = empty_dir;
1565 _make_cpu_key (&key, KEY_FORMAT_3_5, le32_to_cpu (ih->ih_key.k_dir_id),
1566 le32_to_cpu (ih->ih_key.k_objectid), DOT_OFFSET, TYPE_DIRENTRY, 3/*key length*/);
1568 /* compose item head for new item. Directories consist of items of
1569 old type (ITEM_VERSION_1). Do not set key (second arg is 0), it
1570 is done by reiserfs_new_inode */
1571 if (old_format_only (sb)) {
1572 make_le_item_head (ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET, TYPE_DIRENTRY, EMPTY_DIR_SIZE_V1, 2);
1574 make_empty_dir_item_v1 (body, ih->ih_key.k_dir_id, ih->ih_key.k_objectid,
1575 INODE_PKEY (dir)->k_dir_id,
1576 INODE_PKEY (dir)->k_objectid );
1578 make_le_item_head (ih, NULL, KEY_FORMAT_3_5, DOT_OFFSET, TYPE_DIRENTRY, EMPTY_DIR_SIZE, 2);
1580 make_empty_dir_item (body, ih->ih_key.k_dir_id, ih->ih_key.k_objectid,
1581 INODE_PKEY (dir)->k_dir_id,
1582 INODE_PKEY (dir)->k_objectid );
1585 /* look for place in the tree for new item */
1586 retval = search_item (sb, &key, path);
1587 if (retval == IO_ERROR) {
1588 reiserfs_warning (sb, "vs-13080: reiserfs_new_directory: "
1589 "i/o failure occurred creating new directory");
1592 if (retval == ITEM_FOUND) {
1594 reiserfs_warning (sb, "vs-13070: reiserfs_new_directory: "
1595 "object with this key exists (%k)", &(ih->ih_key));
1599 /* insert item, that is empty directory item */
1600 return reiserfs_insert_item (th, path, &key, ih, inode, body);
1604 /* stat data of object has been inserted, this inserts the item
1605 containing the body of symlink */
1606 static int reiserfs_new_symlink (struct reiserfs_transaction_handle *th,
1607 struct inode *inode, /* Inode of symlink */
1608 struct item_head * ih,
1609 struct path * path, const char * symname, int item_len)
1611 struct super_block * sb = th->t_super;
1615 _make_cpu_key (&key, KEY_FORMAT_3_5,
1616 le32_to_cpu (ih->ih_key.k_dir_id),
1617 le32_to_cpu (ih->ih_key.k_objectid),
1618 1, TYPE_DIRECT, 3/*key length*/);
1620 make_le_item_head (ih, NULL, KEY_FORMAT_3_5, 1, TYPE_DIRECT, item_len, 0/*free_space*/);
1622 /* look for place in the tree for new item */
1623 retval = search_item (sb, &key, path);
1624 if (retval == IO_ERROR) {
1625 reiserfs_warning (sb, "vs-13080: reiserfs_new_symlinik: "
1626 "i/o failure occurred creating new symlink");
1629 if (retval == ITEM_FOUND) {
1631 reiserfs_warning (sb, "vs-13080: reiserfs_new_symlink: "
1632 "object with this key exists (%k)", &(ih->ih_key));
1636 /* insert item, that is body of symlink */
1637 return reiserfs_insert_item (th, path, &key, ih, inode, symname);
1641 /* inserts the stat data into the tree, and then calls
1642 reiserfs_new_directory (to insert ".", ".." item if new object is
1643 directory) or reiserfs_new_symlink (to insert symlink body if new
1644 object is symlink) or nothing (if new object is regular file)
1646 NOTE! uid and gid must already be set in the inode. If we return
1647 non-zero due to an error, we have to drop the quota previously allocated
1648 for the fresh inode. This can only be done outside a transaction, so
1649 if we return non-zero, we also end the transaction. */
1650 int reiserfs_new_inode (struct reiserfs_transaction_handle *th,
1651 struct inode * dir, int mode,
1652 const char * symname,
1653 /* 0 for regular, EMTRY_DIR_SIZE for dirs,
1654 strlen (symname) for symlinks)*/
1655 loff_t i_size, struct dentry *dentry,
1656 struct inode *inode)
1658 struct super_block * sb;
1659 INITIALIZE_PATH (path_to_key);
1661 struct item_head ih;
1662 struct stat_data sd;
1666 if (!dir || !dir->i_nlink) {
1673 /* item head of new item */
1674 ih.ih_key.k_dir_id = reiserfs_choose_packing(dir);
1675 ih.ih_key.k_objectid = cpu_to_le32 (reiserfs_get_unused_objectid (th));
1676 if (!ih.ih_key.k_objectid) {
1678 goto out_bad_inode ;
1680 if (old_format_only (sb))
1681 /* not a perfect generation count, as object ids can be reused, but
1682 ** this is as good as reiserfs can do right now.
1683 ** note that the private part of inode isn't filled in yet, we have
1684 ** to use the directory.
1686 inode->i_generation = le32_to_cpu (INODE_PKEY (dir)->k_objectid);
1688 #if defined( USE_INODE_GENERATION_COUNTER )
1689 inode->i_generation = le32_to_cpu(REISERFS_SB(sb)->s_rs->s_inode_generation);
1691 inode->i_generation = ++event;
1694 /* fill stat data */
1695 inode->i_nlink = (S_ISDIR (mode) ? 2 : 1);
1697 /* uid and gid must already be set by the caller for quota init */
1699 /* symlink cannot be immutable or append only, right? */
1700 if( S_ISLNK( inode -> i_mode ) )
1701 inode -> i_flags &= ~ ( S_IMMUTABLE | S_APPEND );
1703 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1704 inode->i_size = i_size;
1705 inode->i_blocks = 0;
1707 REISERFS_I(inode)->i_first_direct_byte = S_ISLNK(mode) ? 1 :
1708 U32_MAX/*NO_BYTES_IN_DIRECT_ITEM*/;
1710 INIT_LIST_HEAD(&(REISERFS_I(inode)->i_prealloc_list ));
1711 REISERFS_I(inode)->i_flags = 0;
1712 REISERFS_I(inode)->i_prealloc_block = 0;
1713 REISERFS_I(inode)->i_prealloc_count = 0;
1714 REISERFS_I(inode)->i_trans_id = 0;
1715 REISERFS_I(inode)->i_jl = NULL;
1716 REISERFS_I(inode)->i_attrs =
1717 REISERFS_I(dir)->i_attrs & REISERFS_INHERIT_MASK;
1718 sd_attrs_to_i_attrs( REISERFS_I(inode) -> i_attrs, inode );
1719 REISERFS_I(inode)->i_acl_access = NULL;
1720 REISERFS_I(inode)->i_acl_default = NULL;
1721 init_rwsem (&REISERFS_I(inode)->xattr_sem);
1723 if (old_format_only (sb))
1724 make_le_item_head (&ih, NULL, KEY_FORMAT_3_5, SD_OFFSET, TYPE_STAT_DATA, SD_V1_SIZE, MAX_US_INT);
1726 make_le_item_head (&ih, NULL, KEY_FORMAT_3_6, SD_OFFSET, TYPE_STAT_DATA, SD_SIZE, MAX_US_INT);
1728 /* key to search for correct place for new stat data */
1729 _make_cpu_key (&key, KEY_FORMAT_3_6, le32_to_cpu (ih.ih_key.k_dir_id),
1730 le32_to_cpu (ih.ih_key.k_objectid), SD_OFFSET, TYPE_STAT_DATA, 3/*key length*/);
1732 /* find proper place for inserting of stat data */
1733 retval = search_item (sb, &key, &path_to_key);
1734 if (retval == IO_ERROR) {
1738 if (retval == ITEM_FOUND) {
1739 pathrelse (&path_to_key);
1743 if (old_format_only (sb)) {
1744 if (inode->i_uid & ~0xffff || inode->i_gid & ~0xffff) {
1745 pathrelse (&path_to_key);
1746 /* i_uid or i_gid is too big to be stored in stat data v3.5 */
1750 inode2sd_v1 (&sd, inode, inode->i_size);
1752 inode2sd (&sd, inode, inode->i_size);
1754 // these do not go to on-disk stat data
1755 inode->i_ino = le32_to_cpu (ih.ih_key.k_objectid);
1756 inode->i_blksize = reiserfs_default_io_size;
1758 // store in in-core inode the key of stat data and version all
1759 // object items will have (directory items will have old offset
1760 // format, other new objects will consist of new items)
1761 memcpy (INODE_PKEY (inode), &(ih.ih_key), KEY_SIZE);
1762 if (old_format_only (sb) || S_ISDIR(mode) || S_ISLNK(mode))
1763 set_inode_item_key_version (inode, KEY_FORMAT_3_5);
1765 set_inode_item_key_version (inode, KEY_FORMAT_3_6);
1766 if (old_format_only (sb))
1767 set_inode_sd_version (inode, STAT_DATA_V1);
1769 set_inode_sd_version (inode, STAT_DATA_V2);
1771 /* insert the stat data into the tree */
1772 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1773 if (REISERFS_I(dir)->new_packing_locality)
1774 th->displace_new_blocks = 1;
1776 retval = reiserfs_insert_item (th, &path_to_key, &key, &ih, inode, (char *)(&sd));
1779 reiserfs_check_path(&path_to_key) ;
1783 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1784 if (!th->displace_new_blocks)
1785 REISERFS_I(dir)->new_packing_locality = 0;
1787 if (S_ISDIR(mode)) {
1788 /* insert item with "." and ".." */
1789 retval = reiserfs_new_directory (th, inode, &ih, &path_to_key, dir);
1792 if (S_ISLNK(mode)) {
1793 /* insert body of symlink */
1794 if (!old_format_only (sb))
1795 i_size = ROUND_UP(i_size);
1796 retval = reiserfs_new_symlink (th, inode, &ih, &path_to_key, symname, i_size);
1800 reiserfs_check_path(&path_to_key) ;
1801 journal_end(th, th->t_super, th->t_blocks_allocated);
1802 goto out_inserted_sd;
1805 /* XXX CHECK THIS */
1806 if (reiserfs_posixacl (inode->i_sb)) {
1807 retval = reiserfs_inherit_default_acl (dir, dentry, inode);
1810 reiserfs_check_path(&path_to_key) ;
1811 journal_end(th, th->t_super, th->t_blocks_allocated);
1812 goto out_inserted_sd;
1814 } else if (inode->i_sb->s_flags & MS_POSIXACL) {
1815 reiserfs_warning (inode->i_sb, "ACLs aren't enabled in the fs, "
1816 "but vfs thinks they are!");
1819 insert_inode_hash (inode);
1820 reiserfs_update_sd(th, inode);
1821 reiserfs_check_path(&path_to_key) ;
1825 /* it looks like you can easily compress these two goto targets into
1826 * one. Keeping it like this doesn't actually hurt anything, and they
1827 * are place holders for what the quota code actually needs.
1830 /* Invalidate the object, nothing was inserted yet */
1831 INODE_PKEY(inode)->k_objectid = 0;
1833 /* dquot_drop must be done outside a transaction */
1834 journal_end(th, th->t_super, th->t_blocks_allocated) ;
1835 DQUOT_FREE_INODE(inode);
1837 inode->i_flags |= S_NOQUOTA;
1838 make_bad_inode(inode);
1842 th->t_trans_id = 0; /* so the caller can't use this handle later */
1848 ** finds the tail page in the page cache,
1849 ** reads the last block in.
1851 ** On success, page_result is set to a locked, pinned page, and bh_result
1852 ** is set to an up to date buffer for the last block in the file. returns 0.
1854 ** tail conversion is not done, so bh_result might not be valid for writing
1855 ** check buffer_mapped(bh_result) and bh_result->b_blocknr != 0 before
1856 ** trying to write the block.
1858 ** on failure, nonzero is returned, page_result and bh_result are untouched.
1860 static int grab_tail_page(struct inode *p_s_inode,
1861 struct page **page_result,
1862 struct buffer_head **bh_result) {
1864 /* we want the page with the last byte in the file,
1865 ** not the page that will hold the next byte for appending
1867 unsigned long index = (p_s_inode->i_size-1) >> PAGE_CACHE_SHIFT ;
1868 unsigned long pos = 0 ;
1869 unsigned long start = 0 ;
1870 unsigned long blocksize = p_s_inode->i_sb->s_blocksize ;
1871 unsigned long offset = (p_s_inode->i_size) & (PAGE_CACHE_SIZE - 1) ;
1872 struct buffer_head *bh ;
1873 struct buffer_head *head ;
1874 struct page * page ;
1877 /* we know that we are only called with inode->i_size > 0.
1878 ** we also know that a file tail can never be as big as a block
1879 ** If i_size % blocksize == 0, our file is currently block aligned
1880 ** and it won't need converting or zeroing after a truncate.
1882 if ((offset & (blocksize - 1)) == 0) {
1885 page = grab_cache_page(p_s_inode->i_mapping, index) ;
1890 /* start within the page of the last block in the file */
1891 start = (offset / blocksize) * blocksize ;
1893 error = block_prepare_write(page, start, offset,
1894 reiserfs_get_block_create_0) ;
1898 head = page_buffers(page) ;
1904 bh = bh->b_this_page ;
1906 } while(bh != head) ;
1908 if (!buffer_uptodate(bh)) {
1909 /* note, this should never happen, prepare_write should
1910 ** be taking care of this for us. If the buffer isn't up to date,
1911 ** I've screwed up the code to find the buffer, or the code to
1912 ** call prepare_write
1914 reiserfs_warning (p_s_inode->i_sb,
1915 "clm-6000: error reading block %lu on dev %s",
1917 reiserfs_bdevname (p_s_inode->i_sb)) ;
1922 *page_result = page ;
1929 page_cache_release(page) ;
1934 ** vfs version of truncate file. Must NOT be called with
1935 ** a transaction already started.
1937 ** some code taken from block_truncate_page
1939 void reiserfs_truncate_file(struct inode *p_s_inode, int update_timestamps) {
1940 struct reiserfs_transaction_handle th ;
1941 /* we want the offset for the first byte after the end of the file */
1942 unsigned long offset = p_s_inode->i_size & (PAGE_CACHE_SIZE - 1) ;
1943 unsigned blocksize = p_s_inode->i_sb->s_blocksize ;
1945 struct page *page = NULL ;
1947 struct buffer_head *bh = NULL ;
1949 reiserfs_write_lock(p_s_inode->i_sb);
1951 if (p_s_inode->i_size > 0) {
1952 if ((error = grab_tail_page(p_s_inode, &page, &bh))) {
1953 // -ENOENT means we truncated past the end of the file,
1954 // and get_block_create_0 could not find a block to read in,
1956 if (error != -ENOENT)
1957 reiserfs_warning (p_s_inode->i_sb,
1958 "clm-6001: grab_tail_page failed %d",
1965 /* so, if page != NULL, we have a buffer head for the offset at
1966 ** the end of the file. if the bh is mapped, and bh->b_blocknr != 0,
1967 ** then we have an unformatted node. Otherwise, we have a direct item,
1968 ** and no zeroing is required on disk. We zero after the truncate,
1969 ** because the truncate might pack the item anyway
1970 ** (it will unmap bh if it packs).
1972 /* it is enough to reserve space in transaction for 2 balancings:
1973 one for "save" link adding and another for the first
1974 cut_from_item. 1 is for update_sd */
1975 journal_begin(&th, p_s_inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1 ) ;
1976 reiserfs_update_inode_transaction(p_s_inode) ;
1977 if (update_timestamps)
1978 /* we are doing real truncate: if the system crashes before the last
1979 transaction of truncating gets committed - on reboot the file
1980 either appears truncated properly or not truncated at all */
1981 add_save_link (&th, p_s_inode, 1);
1982 reiserfs_do_truncate (&th, p_s_inode, page, update_timestamps) ;
1983 journal_end(&th, p_s_inode->i_sb, JOURNAL_PER_BALANCE_CNT * 2 + 1 ) ;
1985 if (update_timestamps)
1986 remove_save_link (p_s_inode, 1/* truncate */);
1989 length = offset & (blocksize - 1) ;
1990 /* if we are not on a block boundary */
1994 length = blocksize - length ;
1995 kaddr = kmap_atomic(page, KM_USER0) ;
1996 memset(kaddr + offset, 0, length) ;
1997 flush_dcache_page(page) ;
1998 kunmap_atomic(kaddr, KM_USER0) ;
1999 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
2000 mark_buffer_dirty(bh) ;
2004 page_cache_release(page) ;
2007 reiserfs_write_unlock(p_s_inode->i_sb);
2010 static int map_block_for_writepage(struct inode *inode,
2011 struct buffer_head *bh_result,
2012 unsigned long block) {
2013 struct reiserfs_transaction_handle th ;
2015 struct item_head tmp_ih ;
2016 struct item_head *ih ;
2017 struct buffer_head *bh ;
2019 struct cpu_key key ;
2020 INITIALIZE_PATH(path) ;
2022 int jbegin_count = JOURNAL_PER_BALANCE_CNT ;
2023 loff_t byte_offset = (block << inode->i_sb->s_blocksize_bits) + 1 ;
2025 int use_get_block = 0 ;
2026 int bytes_copied = 0 ;
2028 int trans_running = 0;
2030 /* catch places below that try to log something without starting a trans */
2033 if (!buffer_uptodate(bh_result)) {
2037 kmap(bh_result->b_page) ;
2039 reiserfs_write_lock(inode->i_sb);
2040 make_cpu_key(&key, inode, byte_offset, TYPE_ANY, 3) ;
2043 retval = search_for_position_by_key(inode->i_sb, &key, &path) ;
2044 if (retval != POSITION_FOUND) {
2049 bh = get_last_bh(&path) ;
2050 ih = get_ih(&path) ;
2051 item = get_item(&path) ;
2052 pos_in_item = path.pos_in_item ;
2054 /* we've found an unformatted node */
2055 if (indirect_item_found(retval, ih)) {
2056 if (bytes_copied > 0) {
2057 reiserfs_warning (inode->i_sb, "clm-6002: bytes_copied %d",
2060 if (!get_block_num(item, pos_in_item)) {
2061 /* crap, we are writing to a hole */
2065 set_block_dev_mapped(bh_result, get_block_num(item,pos_in_item),inode);
2066 } else if (is_direct_le_ih(ih)) {
2068 p = page_address(bh_result->b_page) ;
2069 p += (byte_offset -1) & (PAGE_CACHE_SIZE - 1) ;
2070 copy_size = ih_item_len(ih) - pos_in_item;
2072 fs_gen = get_generation(inode->i_sb) ;
2073 copy_item_head(&tmp_ih, ih) ;
2075 if (!trans_running) {
2076 /* vs-3050 is gone, no need to drop the path */
2077 journal_begin(&th, inode->i_sb, jbegin_count) ;
2078 reiserfs_update_inode_transaction(inode) ;
2080 if (fs_changed(fs_gen, inode->i_sb) && item_moved(&tmp_ih, &path)) {
2081 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
2086 reiserfs_prepare_for_journal(inode->i_sb, bh, 1) ;
2088 if (fs_changed (fs_gen, inode->i_sb) && item_moved (&tmp_ih, &path)) {
2089 reiserfs_restore_prepared_buffer(inode->i_sb, bh) ;
2093 memcpy( B_I_PITEM(bh, ih) + pos_in_item, p + bytes_copied, copy_size) ;
2095 journal_mark_dirty(&th, inode->i_sb, bh) ;
2096 bytes_copied += copy_size ;
2097 set_block_dev_mapped(bh_result, 0, inode);
2099 /* are there still bytes left? */
2100 if (bytes_copied < bh_result->b_size &&
2101 (byte_offset + bytes_copied) < inode->i_size) {
2102 set_cpu_key_k_offset(&key, cpu_key_k_offset(&key) + copy_size) ;
2106 reiserfs_warning (inode->i_sb,
2107 "clm-6003: bad item inode %lu, device %s",
2108 inode->i_ino, reiserfs_bdevname (inode->i_sb)) ;
2116 if (trans_running) {
2117 journal_end(&th, inode->i_sb, jbegin_count) ;
2120 reiserfs_write_unlock(inode->i_sb);
2122 /* this is where we fill in holes in the file. */
2123 if (use_get_block) {
2124 retval = reiserfs_get_block(inode, block, bh_result,
2125 GET_BLOCK_CREATE | GET_BLOCK_NO_ISEM |
2126 GET_BLOCK_NO_DANGLE);
2128 if (!buffer_mapped(bh_result) || bh_result->b_blocknr == 0) {
2129 /* get_block failed to find a mapped unformatted node. */
2135 kunmap(bh_result->b_page) ;
2137 if (!retval && buffer_mapped(bh_result) && bh_result->b_blocknr == 0) {
2138 /* we've copied data from the page into the direct item, so the
2139 * buffer in the page is now clean, mark it to reflect that.
2141 lock_buffer(bh_result);
2142 clear_buffer_dirty(bh_result);
2143 unlock_buffer(bh_result);
2149 * mason@suse.com: updated in 2.5.54 to follow the same general io
2150 * start/recovery path as __block_write_full_page, along with special
2151 * code to handle reiserfs tails.
2153 static int reiserfs_write_full_page(struct page *page, struct writeback_control *wbc) {
2154 struct inode *inode = page->mapping->host ;
2155 unsigned long end_index = inode->i_size >> PAGE_CACHE_SHIFT ;
2157 unsigned long block ;
2158 struct buffer_head *head, *bh;
2161 int checked = PageChecked(page);
2162 struct reiserfs_transaction_handle th;
2163 struct super_block *s = inode->i_sb;
2164 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
2167 /* The page dirty bit is cleared before writepage is called, which
2168 * means we have to tell create_empty_buffers to make dirty buffers
2169 * The page really should be up to date at this point, so tossing
2170 * in the BH_Uptodate is just a sanity check.
2172 if (!page_has_buffers(page)) {
2173 create_empty_buffers(page, s->s_blocksize,
2174 (1 << BH_Dirty) | (1 << BH_Uptodate));
2176 head = page_buffers(page) ;
2178 /* last page in the file, zero out any contents past the
2179 ** last byte in the file
2181 if (page->index >= end_index) {
2183 unsigned last_offset;
2185 last_offset = inode->i_size & (PAGE_CACHE_SIZE - 1) ;
2186 /* no file contents in this page */
2187 if (page->index >= end_index + 1 || !last_offset) {
2191 kaddr = kmap_atomic(page, KM_USER0);
2192 memset(kaddr + last_offset, 0, PAGE_CACHE_SIZE-last_offset) ;
2193 flush_dcache_page(page) ;
2194 kunmap_atomic(kaddr, KM_USER0) ;
2197 block = page->index << (PAGE_CACHE_SHIFT - s->s_blocksize_bits) ;
2198 /* first map all the buffers, logging any direct items we find */
2200 if ((checked || buffer_dirty(bh)) && (!buffer_mapped(bh) ||
2201 (buffer_mapped(bh) && bh->b_blocknr == 0))) {
2202 /* not mapped yet, or it points to a direct item, search
2203 * the btree for the mapping info, and log any direct
2206 if ((error = map_block_for_writepage(inode, bh, block))) {
2210 bh = bh->b_this_page;
2212 } while(bh != head) ;
2215 * we start the transaction after map_block_for_writepage,
2216 * because it can create holes in the file (an unbounded operation).
2217 * starting it here, we can make a reliable estimate for how many
2218 * blocks we're going to log
2221 ClearPageChecked(page);
2222 reiserfs_write_lock(s);
2223 journal_begin(&th, s, bh_per_page + 1);
2224 reiserfs_update_inode_transaction(inode);
2226 /* now go through and lock any dirty buffers on the page */
2229 if (!buffer_mapped(bh))
2231 if (buffer_mapped(bh) && bh->b_blocknr == 0)
2235 reiserfs_prepare_for_journal(s, bh, 1);
2236 journal_mark_dirty(&th, s, bh);
2239 /* from this point on, we know the buffer is mapped to a
2240 * real block and not a direct item
2242 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
2245 if (test_set_buffer_locked(bh)) {
2246 redirty_page_for_writepage(wbc, page);
2250 if (test_clear_buffer_dirty(bh)) {
2251 mark_buffer_async_write(bh);
2255 } while((bh = bh->b_this_page) != head);
2258 journal_end(&th, s, bh_per_page + 1);
2259 reiserfs_write_unlock(s);
2261 BUG_ON(PageWriteback(page));
2262 set_page_writeback(page);
2266 * since any buffer might be the only dirty buffer on the page,
2267 * the first submit_bh can bring the page out of writeback.
2268 * be careful with the buffers.
2271 struct buffer_head *next = bh->b_this_page;
2272 if (buffer_async_write(bh)) {
2273 submit_bh(WRITE, bh);
2278 } while(bh != head);
2284 * if this page only had a direct item, it is very possible for
2285 * no io to be required without there being an error. Or,
2286 * someone else could have locked them and sent them down the
2287 * pipe without locking the page
2291 if (!buffer_uptodate(bh)) {
2295 bh = bh->b_this_page;
2296 } while(bh != head);
2298 SetPageUptodate(page);
2299 end_page_writeback(page);
2304 /* catches various errors, we need to make sure any valid dirty blocks
2305 * get to the media. The page is currently locked and not marked for
2308 ClearPageUptodate(page);
2312 if (buffer_mapped(bh) && buffer_dirty(bh) && bh->b_blocknr) {
2314 mark_buffer_async_write(bh);
2317 * clear any dirty bits that might have come from getting
2318 * attached to a dirty page
2320 clear_buffer_dirty(bh);
2322 bh = bh->b_this_page;
2323 } while(bh != head);
2325 BUG_ON(PageWriteback(page));
2326 set_page_writeback(page);
2329 struct buffer_head *next = bh->b_this_page;
2330 if (buffer_async_write(bh)) {
2331 clear_buffer_dirty(bh);
2332 submit_bh(WRITE, bh);
2337 } while(bh != head);
2342 static int reiserfs_readpage (struct file *f, struct page * page)
2344 return block_read_full_page (page, reiserfs_get_block);
2348 static int reiserfs_writepage (struct page * page, struct writeback_control *wbc)
2350 struct inode *inode = page->mapping->host ;
2351 reiserfs_wait_on_write_block(inode->i_sb) ;
2352 return reiserfs_write_full_page(page, wbc) ;
2355 int reiserfs_prepare_write(struct file *f, struct page *page,
2356 unsigned from, unsigned to) {
2357 struct inode *inode = page->mapping->host ;
2361 reiserfs_wait_on_write_block(inode->i_sb) ;
2362 fix_tail_page_for_writing(page) ;
2363 if (reiserfs_transaction_running(inode->i_sb)) {
2364 struct reiserfs_transaction_handle *th;
2365 th = (struct reiserfs_transaction_handle *)current->journal_info;
2366 old_ref = th->t_refcount;
2370 ret = block_prepare_write(page, from, to, reiserfs_get_block) ;
2371 if (ret && reiserfs_transaction_running(inode->i_sb)) {
2372 struct reiserfs_transaction_handle *th = current->journal_info;
2373 /* this gets a little ugly. If reiserfs_get_block returned an
2374 * error and left a transacstion running, we've got to close it,
2375 * and we've got to free handle if it was a persistent transaction.
2377 * But, if we had nested into an existing transaction, we need
2378 * to just drop the ref count on the handle.
2380 * If old_ref == 0, the transaction is from reiserfs_get_block,
2381 * and it was a persistent trans. Otherwise, it was nested above.
2383 if (th->t_refcount > old_ref) {
2387 reiserfs_write_lock(inode->i_sb);
2388 reiserfs_end_persistent_transaction(th);
2389 reiserfs_write_unlock(inode->i_sb);
2398 static sector_t reiserfs_aop_bmap(struct address_space *as, sector_t block) {
2399 return generic_block_bmap(as, block, reiserfs_bmap) ;
2402 static int reiserfs_commit_write(struct file *f, struct page *page,
2403 unsigned from, unsigned to) {
2404 struct inode *inode = page->mapping->host ;
2405 loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
2408 struct reiserfs_transaction_handle *th = NULL;
2410 reiserfs_wait_on_write_block(inode->i_sb) ;
2411 if (reiserfs_transaction_running(inode->i_sb)) {
2412 th = current->journal_info;
2414 reiserfs_commit_page(inode, page, from, to);
2416 /* generic_commit_write does this for us, but does not update the
2417 ** transaction tracking stuff when the size changes. So, we have
2418 ** to do the i_size updates here.
2420 if (pos > inode->i_size) {
2421 struct reiserfs_transaction_handle myth ;
2422 reiserfs_write_lock(inode->i_sb);
2423 /* If the file have grown beyond the border where it
2424 can have a tail, unmark it as needing a tail
2426 if ( (have_large_tails (inode->i_sb) && inode->i_size > i_block_size (inode)*4) ||
2427 (have_small_tails (inode->i_sb) && inode->i_size > i_block_size(inode)) )
2428 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask ;
2430 journal_begin(&myth, inode->i_sb, 1) ;
2431 reiserfs_update_inode_transaction(inode) ;
2432 inode->i_size = pos ;
2433 reiserfs_update_sd(&myth, inode) ;
2435 journal_end(&myth, inode->i_sb, 1) ;
2436 reiserfs_write_unlock(inode->i_sb);
2439 reiserfs_write_lock(inode->i_sb);
2441 reiserfs_update_sd(th, inode) ;
2442 reiserfs_end_persistent_transaction(th);
2443 reiserfs_write_unlock(inode->i_sb);
2446 /* we test for O_SYNC here so we can commit the transaction
2447 ** for any packed tails the file might have had
2449 if (f && (f->f_flags & O_SYNC)) {
2450 reiserfs_write_lock(inode->i_sb);
2451 reiserfs_commit_for_inode(inode) ;
2452 reiserfs_write_unlock(inode->i_sb);
2457 void sd_attrs_to_i_attrs( __u16 sd_attrs, struct inode *inode )
2459 if( reiserfs_attrs( inode -> i_sb ) ) {
2460 if( sd_attrs & REISERFS_SYNC_FL )
2461 inode -> i_flags |= S_SYNC;
2463 inode -> i_flags &= ~S_SYNC;
2464 if( sd_attrs & REISERFS_IMMUTABLE_FL )
2465 inode -> i_flags |= S_IMMUTABLE;
2467 inode -> i_flags &= ~S_IMMUTABLE;
2468 if( sd_attrs & REISERFS_IUNLINK_FL )
2469 inode -> i_flags |= S_IUNLINK;
2471 inode -> i_flags &= ~S_IUNLINK;
2472 if( sd_attrs & REISERFS_BARRIER_FL )
2473 inode -> i_flags |= S_BARRIER;
2475 inode -> i_flags &= ~S_BARRIER;
2476 if( sd_attrs & REISERFS_APPEND_FL )
2477 inode -> i_flags |= S_APPEND;
2479 inode -> i_flags &= ~S_APPEND;
2480 if( sd_attrs & REISERFS_NOATIME_FL )
2481 inode -> i_flags |= S_NOATIME;
2483 inode -> i_flags &= ~S_NOATIME;
2484 if( sd_attrs & REISERFS_NOTAIL_FL )
2485 REISERFS_I(inode)->i_flags |= i_nopack_mask;
2487 REISERFS_I(inode)->i_flags &= ~i_nopack_mask;
2491 void i_attrs_to_sd_attrs( struct inode *inode, __u16 *sd_attrs )
2493 if( reiserfs_attrs( inode -> i_sb ) ) {
2494 if( inode -> i_flags & S_IMMUTABLE )
2495 *sd_attrs |= REISERFS_IMMUTABLE_FL;
2497 *sd_attrs &= ~REISERFS_IMMUTABLE_FL;
2498 if( inode -> i_flags & S_IUNLINK )
2499 *sd_attrs |= REISERFS_IUNLINK_FL;
2501 *sd_attrs &= ~REISERFS_IUNLINK_FL;
2502 if( inode -> i_flags & S_BARRIER )
2503 *sd_attrs |= REISERFS_BARRIER_FL;
2505 *sd_attrs &= ~REISERFS_BARRIER_FL;
2506 if( inode -> i_flags & S_SYNC )
2507 *sd_attrs |= REISERFS_SYNC_FL;
2509 *sd_attrs &= ~REISERFS_SYNC_FL;
2510 if( inode -> i_flags & S_NOATIME )
2511 *sd_attrs |= REISERFS_NOATIME_FL;
2513 *sd_attrs &= ~REISERFS_NOATIME_FL;
2514 if( REISERFS_I(inode)->i_flags & i_nopack_mask )
2515 *sd_attrs |= REISERFS_NOTAIL_FL;
2517 *sd_attrs &= ~REISERFS_NOTAIL_FL;
2521 /* decide if this buffer needs to stay around for data logging or ordered
2524 static int invalidatepage_can_drop(struct inode *inode, struct buffer_head *bh)
2527 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb) ;
2529 spin_lock(&j->j_dirty_buffers_lock) ;
2530 if (!buffer_mapped(bh)) {
2533 /* the page is locked, and the only places that log a data buffer
2534 * also lock the page.
2536 if (reiserfs_file_data_log(inode)) {
2538 * very conservative, leave the buffer pinned if
2539 * anyone might need it.
2541 if (buffer_journaled(bh) || buffer_journal_dirty(bh)) {
2545 if (buffer_dirty(bh) || buffer_locked(bh)) {
2546 struct reiserfs_journal_list *jl;
2547 struct reiserfs_jh *jh = bh->b_private;
2549 /* why is this safe?
2550 * reiserfs_setattr updates i_size in the on disk
2551 * stat data before allowing vmtruncate to be called.
2553 * If buffer was put onto the ordered list for this
2554 * transaction, we know for sure either this transaction
2555 * or an older one already has updated i_size on disk,
2556 * and this ordered data won't be referenced in the file
2559 * if the buffer was put onto the ordered list for an older
2560 * transaction, we need to leave it around
2562 if (jh && (jl = jh->jl) && jl != SB_JOURNAL(inode->i_sb)->j_current_jl)
2566 if (ret && bh->b_private) {
2567 reiserfs_free_jh(bh);
2569 spin_unlock(&j->j_dirty_buffers_lock) ;
2573 /* clm -- taken from fs/buffer.c:block_invalidate_page */
2574 static int reiserfs_invalidatepage(struct page *page, unsigned long offset)
2576 struct buffer_head *head, *bh, *next;
2577 struct inode *inode = page->mapping->host;
2578 unsigned int curr_off = 0;
2581 BUG_ON(!PageLocked(page));
2584 ClearPageChecked(page);
2586 if (!page_has_buffers(page))
2589 head = page_buffers(page);
2592 unsigned int next_off = curr_off + bh->b_size;
2593 next = bh->b_this_page;
2596 * is this block fully invalidated?
2598 if (offset <= curr_off) {
2599 if (invalidatepage_can_drop(inode, bh))
2600 reiserfs_unmap_buffer(bh);
2604 curr_off = next_off;
2606 } while (bh != head);
2609 * We release buffers only if the entire page is being invalidated.
2610 * The get_block cached value has been unconditionally invalidated,
2611 * so real IO is not possible anymore.
2614 ret = try_to_release_page(page, 0);
2619 static int reiserfs_set_page_dirty(struct page *page) {
2620 struct inode *inode = page->mapping->host;
2621 if (reiserfs_file_data_log(inode)) {
2622 SetPageChecked(page);
2623 return __set_page_dirty_nobuffers(page);
2625 return __set_page_dirty_buffers(page);
2629 * Returns 1 if the page's buffers were dropped. The page is locked.
2631 * Takes j_dirty_buffers_lock to protect the b_assoc_buffers list_heads
2632 * in the buffers at page_buffers(page).
2634 * even in -o notail mode, we can't be sure an old mount without -o notail
2635 * didn't create files with tails.
2637 static int reiserfs_releasepage(struct page *page, int unused_gfp_flags)
2639 struct inode *inode = page->mapping->host ;
2640 struct reiserfs_journal *j = SB_JOURNAL(inode->i_sb) ;
2641 struct buffer_head *head ;
2642 struct buffer_head *bh ;
2645 WARN_ON(PageChecked(page));
2646 spin_lock(&j->j_dirty_buffers_lock) ;
2647 head = page_buffers(page) ;
2650 if (bh->b_private) {
2651 if (!buffer_dirty(bh) && !buffer_locked(bh)) {
2652 reiserfs_free_jh(bh);
2658 bh = bh->b_this_page ;
2659 } while (bh != head) ;
2661 ret = try_to_free_buffers(page) ;
2662 spin_unlock(&j->j_dirty_buffers_lock) ;
2666 /* We thank Mingming Cao for helping us understand in great detail what
2667 to do in this section of the code. */
2668 static ssize_t reiserfs_direct_IO(int rw, struct kiocb *iocb,
2669 const struct iovec *iov, loff_t offset, unsigned long nr_segs)
2671 struct file *file = iocb->ki_filp;
2672 struct inode *inode = file->f_mapping->host;
2674 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2675 offset, nr_segs, reiserfs_get_blocks_direct_io, NULL);
2678 int reiserfs_setattr_flags(struct inode *inode, unsigned int flags)
2680 unsigned int oldflags, newflags;
2682 oldflags = REISERFS_I(inode)->i_flags;
2683 newflags = oldflags & ~(REISERFS_IMMUTABLE_FL |
2684 REISERFS_IUNLINK_FL | REISERFS_BARRIER_FL);
2685 if (flags & ATTR_FLAG_IMMUTABLE)
2686 newflags |= REISERFS_IMMUTABLE_FL;
2687 if (flags & ATTR_FLAG_IUNLINK)
2688 newflags |= REISERFS_IUNLINK_FL;
2689 if (flags & ATTR_FLAG_BARRIER)
2690 newflags |= REISERFS_BARRIER_FL;
2692 if (oldflags ^ newflags) {
2693 REISERFS_I(inode)->i_flags = newflags;
2694 inode->i_ctime = CURRENT_TIME;
2699 int reiserfs_setattr(struct dentry *dentry, struct iattr *attr) {
2700 struct inode *inode = dentry->d_inode ;
2702 unsigned int ia_valid = attr->ia_valid;
2704 reiserfs_write_lock(inode->i_sb);
2705 if (S_ISDIR(inode->i_mode))
2708 if (attr->ia_valid & ATTR_SIZE) {
2709 /* version 2 items will be caught by the s_maxbytes check
2710 ** done for us in vmtruncate
2712 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5 &&
2713 attr->ia_size > MAX_NON_LFS) {
2717 /* fill in hole pointers in the expanding truncate case. */
2718 if (attr->ia_size > inode->i_size) {
2719 error = generic_cont_expand(inode, attr->ia_size) ;
2720 if (REISERFS_I(inode)->i_prealloc_count > 0) {
2721 struct reiserfs_transaction_handle th ;
2722 /* we're changing at most 2 bitmaps, inode + super */
2723 journal_begin(&th, inode->i_sb, 4) ;
2724 reiserfs_discard_prealloc (&th, inode);
2725 journal_end(&th, inode->i_sb, 4) ;
2732 if ((((attr->ia_valid & ATTR_UID) && (attr->ia_uid & ~0xffff)) ||
2733 ((attr->ia_valid & ATTR_GID) && (attr->ia_gid & ~0xffff))) &&
2734 (get_inode_sd_version (inode) == STAT_DATA_V1)) {
2735 /* stat data of format v3.5 has 16 bit uid and gid */
2741 error = inode_change_ok(inode, attr) ;
2743 if (!error && attr->ia_valid & ATTR_ATTR_FLAG)
2744 reiserfs_setattr_flags(inode, attr->ia_attr_flags);
2747 if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
2748 (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid)) {
2749 error = reiserfs_chown_xattrs (inode, attr);
2752 error = DQUOT_TRANSFER(inode, attr) ? -EDQUOT : 0;
2755 error = inode_setattr(inode, attr) ;
2759 if (!error && reiserfs_posixacl (inode->i_sb)) {
2760 if (attr->ia_valid & ATTR_MODE)
2761 error = reiserfs_acl_chmod (inode);
2765 reiserfs_write_unlock(inode->i_sb);
2771 struct address_space_operations reiserfs_address_space_operations = {
2772 .writepage = reiserfs_writepage,
2773 .readpage = reiserfs_readpage,
2774 .readpages = reiserfs_readpages,
2775 .releasepage = reiserfs_releasepage,
2776 .invalidatepage = reiserfs_invalidatepage,
2777 .sync_page = block_sync_page,
2778 .prepare_write = reiserfs_prepare_write,
2779 .commit_write = reiserfs_commit_write,
2780 .bmap = reiserfs_aop_bmap,
2781 .direct_IO = reiserfs_direct_IO,
2782 .set_page_dirty = reiserfs_set_page_dirty,